Anyone here besides me read "2010" (the second book) by Arthur C. Clarke?
In it, Clarke explains the reason HAL went nuts, basically the same way Ben did about Petey.
Computers, especially those with higher functioning than the "ordinary" ones, can, and have, gone insane when they have unresovable logical dillemas.
(In HAL's case, his original programming was to process information accurately, but he was told to lie to the "awake" crew.)
I read about one computer, in a car factory that controlled a number of robots making cars, when it was told to "speed things up" by a CEO beyond it's working parameters went crazy, and it's robots destroyed a number of half-finished cars.
Ralph, did you base your view of Petey's insanity on the premise Clarke made about HAL?
Petey=Hal?
I'm not aware of the car company situation you mention, but being lightly familiar with that sort of embedded controller system, the anlogy doesn't work.
The "robots" building cars and assembling chassis are not "aware", in the classic HAL sense. They're no more "aware" than your desktop PC- in fact, most of the assembly systems are run by off-the-shelf PC hardware including networking protocols. Specialized programming, of course, but standard hardware.
They're given certain instructions ("X number of spotwelds along this seam") and postition sensors so that it "knows" where both the arm and welder are, and the seam is (or is supposed to be.)
I doubt the CEO just went to a convenient panel and typed in "make go faster" or the equivalent. Besides the security of the system- like any good IT network, only a few people have the admin power to make anything over cursory changes- there's things like conveyor speed, weld cycle times, operator cycle times (how fast the crewman can get the panel in place) and assembly speeds for the other components. (It makes little sense to be able to make 20 bodies in an hour if you can only make 16 rolling chassis in an hour.)
Clarke hinted at the cause of HALs psychosis even in the first book, but took pains to explain it in the second.
The "Ayn Rand" circuit.
Clever.
Rennen
The "robots" building cars and assembling chassis are not "aware", in the classic HAL sense. They're no more "aware" than your desktop PC- in fact, most of the assembly systems are run by off-the-shelf PC hardware including networking protocols. Specialized programming, of course, but standard hardware.
They're given certain instructions ("X number of spotwelds along this seam") and postition sensors so that it "knows" where both the arm and welder are, and the seam is (or is supposed to be.)
I doubt the CEO just went to a convenient panel and typed in "make go faster" or the equivalent. Besides the security of the system- like any good IT network, only a few people have the admin power to make anything over cursory changes- there's things like conveyor speed, weld cycle times, operator cycle times (how fast the crewman can get the panel in place) and assembly speeds for the other components. (It makes little sense to be able to make 20 bodies in an hour if you can only make 16 rolling chassis in an hour.)
Clarke hinted at the cause of HALs psychosis even in the first book, but took pains to explain it in the second.
The "Ayn Rand" circuit.
Clever.Rennen
-
ZOMBIE USER 12759
- Regular Poster
- Posts: 53
- Joined: Sun Feb 27, 2005 2:26 am
I've heard of an incident like that. The robots didn't go crazy or rebel or anything, they just went on the fritz. I don't remember if they actually cut up a car chassis, but it's not inconceivable that the malfuntion caused a cutting arm to destroy something. Of course, it's not like the robot "Decided" to do anything. More like a short circuit turning on your headlights when it rains. (or your horn, that happened to me a while ago)
Robots
Well, let's see now.
If I remember it correctly, this happened in Japan.
The robots were controlled by a central computer.
Some robots that are used in manufacturing cars are rather large, due to handling large chunks of of the car regularly.
If the robots in question "go wild," they have the capability of doing major damage before they can be shut down.
(I read of one industrial robot that smashed in someone's skull.)
I personally saw a robot at a computer convention go completely out of control, and run right into a monitor, destroying it.
And said robot went out of control due to stray radio signals, probably from the police station a block from the convention.
And then of course there are those robots in TV shows like "Robot Wars".....
If I remember it correctly, this happened in Japan.
The robots were controlled by a central computer.
Some robots that are used in manufacturing cars are rather large, due to handling large chunks of of the car regularly.
If the robots in question "go wild," they have the capability of doing major damage before they can be shut down.
(I read of one industrial robot that smashed in someone's skull.)
I personally saw a robot at a computer convention go completely out of control, and run right into a monitor, destroying it.
And said robot went out of control due to stray radio signals, probably from the police station a block from the convention.
And then of course there are those robots in TV shows like "Robot Wars".....
Well, linked to a central computer, anyway. Much like an offices' IT setup, the individual systems have their own capabilties, but are or can be linked with the network for data comparison and feedback.
And there's "going wild" with bad or corrupted "read only" programming, and "going wild" with a feral AI.
Happens to us in the CNC machining world every day. Like programming any other computer, if you so much as put a semicolon in in place of a colon, well, that might tell your toolholder to head back to "reset" instead of continue the cut for another twenty millimeters.
Except when it "heads back" it tries to push the tool through the wall of the void just milled, wrecking the part, the tombstone (the part holder) the tool itself and usually damaging the toolhead as well.
The more input a system gets, the more likely for an out-of-parameter failure. In the automotive line, there's position sensors to tell the system if the welding arm is "stowed", if the welder is 'hot', if the next body is in position and so on, then there's dynamic sensors that actually read the current, real-time position of the seam to be welded, etc.
Ever had your PC "lock up"? Same thing happens to an assembler- there's a hundred thousand lines of dynamic code, and every now and then a parameter falls out of place and all of a sudden the arm thinks it's 180 degrees out and is now swinging around in the wrong line, trying to spot-weld thin air.
If some guy is unlucky enough to be standing there, yeah, that arm can klunk him a good one. But that doesn't mean the 'Bot "went nuts" and "attacked" the guy.
And the 'Bots in Robot Wars and Battlebots are by no means actually robots. More accurately, they're large, fancy radio-controlled cars. They're not the least bit autonomous, and none of them that I know of have even rudimentary sensors (other than maybe positioning encoders, so the drivers' controller can indicate things like the position of a lifting arm, etc.)
A Battlebot "going wild" means, for example, the contactors "welded" on the weapon relay- since the motor is drawing 200 amps at 48 volts, say- and so they can't safely shut down the weapon. In which case they just wait 'til the batteries die- most 'bots are sucking the most power they possibly can from impressive arrays of NiCads, but a 15-horsepower Etek motor will still draw 'em flat in about four to six minutes.
There's no such thing as a Battlebot going "wild" and attacking someone- there's no controller, program or software, no sensors, no nothing.
That's not to say... ahem.... that people don't get hurt running or building them, but that doesn't mean they're going to leap off the workbench and start chasing the dog.
It's late and I've written too much already.
Rennen
And there's "going wild" with bad or corrupted "read only" programming, and "going wild" with a feral AI.
Happens to us in the CNC machining world every day. Like programming any other computer, if you so much as put a semicolon in in place of a colon, well, that might tell your toolholder to head back to "reset" instead of continue the cut for another twenty millimeters.
Except when it "heads back" it tries to push the tool through the wall of the void just milled, wrecking the part, the tombstone (the part holder) the tool itself and usually damaging the toolhead as well.
The more input a system gets, the more likely for an out-of-parameter failure. In the automotive line, there's position sensors to tell the system if the welding arm is "stowed", if the welder is 'hot', if the next body is in position and so on, then there's dynamic sensors that actually read the current, real-time position of the seam to be welded, etc.
Ever had your PC "lock up"? Same thing happens to an assembler- there's a hundred thousand lines of dynamic code, and every now and then a parameter falls out of place and all of a sudden the arm thinks it's 180 degrees out and is now swinging around in the wrong line, trying to spot-weld thin air.
If some guy is unlucky enough to be standing there, yeah, that arm can klunk him a good one. But that doesn't mean the 'Bot "went nuts" and "attacked" the guy.
And the 'Bots in Robot Wars and Battlebots are by no means actually robots. More accurately, they're large, fancy radio-controlled cars. They're not the least bit autonomous, and none of them that I know of have even rudimentary sensors (other than maybe positioning encoders, so the drivers' controller can indicate things like the position of a lifting arm, etc.)
A Battlebot "going wild" means, for example, the contactors "welded" on the weapon relay- since the motor is drawing 200 amps at 48 volts, say- and so they can't safely shut down the weapon. In which case they just wait 'til the batteries die- most 'bots are sucking the most power they possibly can from impressive arrays of NiCads, but a 15-horsepower Etek motor will still draw 'em flat in about four to six minutes.
There's no such thing as a Battlebot going "wild" and attacking someone- there's no controller, program or software, no sensors, no nothing.
That's not to say... ahem.... that people don't get hurt running or building them, but that doesn't mean they're going to leap off the workbench and start chasing the dog.
It's late and I've written too much already.
Rennen